Gene Set Enrichment Analysis Genome 559: Introduction to Statistical and Computational Genomics Elhanan Borenstein

 Gene expression profiling  Which molecular processes/functions are involved in a certain phenotype (e.g., disease, stress response, etc.)  The Gene Ontology (GO) Project  Provides shared vocabulary/annotation  GO terms are linked in a complex structure  Enrichment analysis:  Find the “most” differentially expressed genes  Identify functional annotations that are over-represented  Modified Fisher's exact test A quick review

A quick review: Modified Fisher's exact test Differentially expressed (DE) genes/balls 4 out of 810 out of 50 2 out of 8 4 out of 81 out of 82 out of 85 out of 83 out of 8 Null model: the 8 genes/balls are selected randomly … So, if you have 50 balls, 10 of them are blue, and you pick 8 balls randomly, what is the probability that k of them are blue? Do I have a surprisingly high number of blue genes? Genes/balls

A quick review: Modified Fisher's exact test m=50, m t =10, n=8 Hypergeometric distribution So … do I have a surprisingly high number of blue genes? Can such high numbers (4 or above) occur by change? What is the probability of getting at least 4 blue genes in the null model? P(σ t >=4) Probability k

Enrichment Analysis ClassA ClassB Genes ranked by expression correlation to Class A Cutoff Biological function?

Enrichment Analysis ClassA ClassB Genes ranked by expression correlation to Class A Cutoff Biological function? 2 / 10 Function 1 (e.g., metabolism) 5 / 11 Function 2 (e.g., signaling) 3 / 10 Function 3 (e.g., regulation)

 After correcting for multiple hypotheses testing, no individual gene may meet the threshold due to noise.  Alternatively, one may be left with a long list of significant genes without any unifying biological theme.  The cutoff value is often arbitrary!  We are really examining only a handful of genes, totally ignoring much of the data Problems with cutoff-based analysis

 MIT, Broad Institute  V 2.0 available since Jan 2007 Gene Set Enrichment Analysis (Subramanian et al. PNAS )

 Calculates a score for the enrichment of a entire set of genes rather than single genes!  Does not require setting a cutoff!  Identifies the set of relevant genes as part of the analysis!  Provides a more robust statistical framework! GSEA key features

Gene Set Enrichment Analysis ClassA ClassB Genes ranked by expression correlation to Class A Cutoff Biological function? 2 / 10 5 / 11 3 / 10 Function 1 (e.g., metabolism) Function 2 (e.g., signaling) Function 3 (e.g., regulation)

Gene Set Enrichment Analysis ClassA ClassB Genes ranked by expression correlation to Class A Running sum: Increase when gene is in set Decrease otherwise Function 1 (e.g., metabolism) Function 2 (e.g., signaling) Function 3 (e.g., regulation)

Gene Set Enrichment Analysis What would you expect if the hits were randomly distributed? What would you expect if most of the hits cluster at the top of the list?

Gene Set Enrichment Analysis Genes within functional set (hits) Running sum Enrichment score (ES) = max deviation from 0 Leading Edge genes

Gene Set Enrichment Analysis Low ES (evenly distributed) ES = 0.43ES = -0.45

Ducray et al. Molecular Cancer :41 Gene Set Enrichment Analysis

1.Calculation of an enrichment score (ES) for each functional category 2.Estimation of significance level of the ES  An empirical permutation test  Phenotype labels are shuffled and the ES for this functional set is recomputed. Repeat 1000 times.  Generating a null distribution 3.Adjustment for multiple hypotheses testing  Necessary if comparing multiple gene sets (i.e.,functions)  Computes FDR (false discovery rate) GSEA Steps